The Download: computing’s next leap and permafrost’s climate threat

In 1959, physicist Richard Feynman argued that computing could shrink drastically, suggesting that the entire Encyclopædia Britannica could one day fit on the head of a pin. For decades, this vision became reality as transistors—tiny switches that power computers—shrank by half every few years, driving faster and smaller devices. However, this progress has slowed because transistors are now so tiny that physical laws limit further miniaturization. Today, the computing industry is turning to brute force solutions, such as building enormous data centers packed with servers to boost performance. Despite this shift, researchers believe there is still untapped potential in computing’s future. They are exploring innovative approaches to overcome current limitations, including new materials, quantum computing, and alternative architectures to unlock the next major leap in performance.

Permafrost is a layer of frozen soil found in polar and high-altitude regions, such as Alaska’s forests. It stores vast amounts of carbon, which has been locked away for thousands of years. As global temperatures rise, permafrost is thawing, releasing stored carbon into the atmosphere as carbon dioxide and methane—two potent greenhouse gases. This process creates a feedback loop: thawing permafrost accelerates climate change, which in turn causes more thawing. Recent reports highlight the growing threat of wildfires in these regions, which further destabilize permafrost and release additional greenhouse gases. Understanding this cycle is critical for predicting future climate impacts and developing mitigation strategies.

In 2025, Samuel King, a PhD student at Stanford University, used a generative AI model—a type of artificial intelligence that creates new content based on patterns in data—to design genetic blueprints for microscopic viruses. This work does not yet represent AI-generated life, as the viruses have not been brought to life. However, it marks a significant step toward using AI to explore biology and potentially engineer new life forms. King’s research was recognized when he was named one of MIT Technology Review’s Innovators Under 35. The project raises ethical questions about the boundaries of AI in science and the potential risks of creating novel biological entities.

In a recent discovery, AI identified a decade-old flaw in PHP, a widely used programming language that powers many websites. The flaw posed a significant security risk, as it could allow hackers to exploit vulnerabilities in countless sites. Governments have urged website owners to patch PHP by updating their software or face potential shutdowns. Since the disclosure, attempts to exploit the flaw have surged from thousands to millions. This incident highlights the growing role of AI in cybersecurity, both as a tool for identifying threats and as a potential weapon for malicious actors. It underscores the need for proactive measures to protect digital infrastructure.

Hackers have used advanced AI agents—autonomous programs capable of performing tasks independently—to automate attacks on over 400 banks in Latin America. These AI agents exploited vulnerabilities in financial systems, demonstrating how AI can be weaponized for cybercrime. The incidents have raised concerns about the security of AI tools, particularly those that operate without human oversight. Meanwhile, financial institutions like Barclays are exploring AI to cut costs, including reducing their workforce by 5,000 jobs. This dual use of AI—both as a tool for innovation and a vector for crime—poses challenges for regulators and security experts.

The Pentagon has authorized the Department of Homeland Security (DHS) to test a heat ray weapon, designed for crowd control at the southern U.S. border. The weapon uses invisible energy beams to create a burning sensation on the skin, allowing authorities to disperse crowds without physical contact. This technology is part of a broader trend toward non-lethal weapons in law enforcement and military operations. The testing reflects ongoing efforts to develop advanced tools for managing large gatherings or potential threats while minimizing harm.

A startup claims to have used AI-designed viruses to treat a patient for the first time. The experimental treatment targeted drug-resistant E. coli, a type of bacteria that does not respond to conventional antibiotics. While the company reports success, it has not yet published peer-reviewed results—findings validated by independent experts—which are essential for scientific credibility. AI-designed viruses are already being explored as a tool to combat bacterial infections, offering a potential alternative to antibiotics. However, the lack of published data raises questions about the treatment’s safety and efficacy.

Japan’s space agency is attempting to return the first samples from Phobos, one of Mars’s moons, using its spacecraft. The mission aims to determine whether Mars was once habitable by analyzing the composition of Phobos, which may contain material ejected from the planet’s surface. This effort could provide insights into the history of Mars and its potential to support life. The spacecraft is currently en route to Phobos, marking a significant milestone in space exploration and international efforts to study the solar system.

Meta, the parent company of Facebook, faces a potential penalty of up to $25 billion over its role in the Cambridge Analytica scandal. The scandal involved the unauthorized harvesting of personal data from millions of Facebook users, which was then used for political advertising. Meta has announced plans to appeal the ruling issued in New Mexico, arguing against the penalty’s validity. The case highlights ongoing concerns about data privacy, corporate accountability, and the ethical use of user data in the digital age.